About

MagPunk is not a regular RSS reader. It is a content analysis and intelligence engine running on your own device, equipped with local machine learning models. We have no servers, no user accounts, and no data mining; all control and power are in your hands.

Who is it for?

• Information professionals and researchers: Those who want to follow the agenda at its source, measure the spread speed of content and the reliability of sources.
• Data privacy advocates: Those who refuse the profiling of their reading habits and seek zero-telemetry.
• Those who want to focus: Those tired of algorithmic timelines and clickbait, who want to do their own curation.

What does it do?

It classifies content on your device; groups sources reporting the same event and rewards the pioneer source that provides the content first, catches fast-spreading events, and moves important developments to the top of the feed. It presents critical content in the Pulse feed throughout the day, and as an offline Daily Brief in the evenings. You can see the reason for every label it assigns word by word.

The details of these mechanisms are laid out across 57 cards in Understand MagPunk.

Our transparency promise

Except for fetching content and the optionally enabled Smart Discovery, no data ever leaves your device. Your watched words, reading history, and all your analyses belong solely to you. In the name of transparency, we publicly share the classification model we use, its weight tables and the verification data behind our report card on the Data & Transparency page — including the rows where the model is weak.

How the data sits on your device is written out on the Privacy Policy page; what the labels claim and what they do not is on the Automatic Classification page.

Who builds it?

MagPunk Labs — Heval Deniz Kaya.

It is a small operation and is meant to stay that way: with no server, there is neither running cost nor accumulated data about you. Everything needed for daily use is free; an optional Pro subscription unlocks advanced features and keeps the project ad-free.

You can reach us through the channels on the Contact page.